How does current AI for technological innovation work

[Shagnasty]

@RaftPeople

Shagnasty, while I wouldn’t disagree that much of what you say is correct, you seem to be forgetting the substantial progress made with neural networks. Unlike algorithmic programming (where you get comma vs semicolon type problems), neural networks have been used successfully to solve problems where the input information is not always complete, and it avoids the problem of having to anticipate in the program every path that is possible.

As for computing all possibilities subconsciously: I think that can be ruled out with math, there just isn’t enough time for our brain to do that for complex problems.

I think you are not giving enough credit to those areas where there has been much success, speech recognition, face recognition, spotting patterns in financial markets, etc. Neural networks are clearly far simpler than what is going on in the brain, but they are understood by mathemeticians (to a degree), and they perform orders of magnitude better than traditional programming.

I am the biggest computer fan there is. I also understand the brain better than most people. I am not trying to downplay the acheivements of either one.

It is just that things like voice and face recognition were once thought to be failry easy problems and it turns out that they are extremely difficult.

Neural networks hold some promise as well. I am just saying that AI lagged way behind where most people thought it would be while other apsects of computing surged ahead.

Computers are just tools and at work, I guide some huge ones along like a farmer would a giant combine. That is great but it isn’t like the combine is going to suggest new ways to fertilize the fields.

[Wesley_Clark]

@Shagnasty

I suppose it is possible that the brain is computing all alternatives subconsciencly.

There is truth to that.

http://www.futurepundit.com/archives/cat_brain_enhancement.html

One group was given four minutes to pick a favourite car from a list having weighed up four attributes including fuel consumption and legroom.

The other group was given a series of puzzles to keep their conscious selves busy before making a decision.

The conscious thought group managed to pick the best car based on four aspects around 55% of the time, while the unconscious thought group only chose the right one 40% of the time.

But when the experiment was made more complex by bringing in 12 attributes to weigh up, the conscious thought group’s success rate fell to around 23% as opposed to nearly 60% for the unconscious thought group.

[Voyager]

@Wesley Clark

My brother also told me that computer designing comes up with circuit board designs far better than anything designed by humans. I have also read that humans only catch 98% of spam emails when exposed to the subject heading, but computer programs catch closer to 99.5%. So as it stands it seems computers can design some things more effectively than humans. How efficient and how pervasive is this kind of soft AI in 2006?

I’ve been using and designing software to help design computers for 25 years. Your brother is correct - circuit board design couldn’t be done without computer help. (At least not in reasonable time.) Routing algorithms are much better than people, and they can take things like signal integrity into account also. The same goes for routing lines inside a chip. I’m a co-inventor on a routing patent, mostly because I threw in a good idea to the person who did the real work.

But these programs use heuristics and aren’t AI in the slightest. Some of the heuristics did get developed as a spinoff of AI research, but while they help solve problems they don’t simulate the way we do it. Real AI-like techniques for big problems, like GAs, aren’t nearly as efficient as purely algorithmic and heuristic techniques.

The best way of looking at this is that the programs are tools that allow the people to concentrate on the creative parts of the job, not the mechanical ones.

I don’t buy any dates for AI to work, since they don’t seem to be any closer to understanding human thought than they were when I took AI classes in 1972. Back then they thought it would be all ready in 30 years also.

While this may be GD territory, beware those who think that more powerful computers always lead to more sophisticated software. The code might handle bigger jobs, and run faster, but it doesn’t write itself. I was in a hotel when the 386 came out, and the story in USA Today said that with real 16 bit computing the AI problem would be solved in no time. So, a lot of nonsense has been written on this issue.